w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Review

Activated sludge rheology: A critical review on data collection

and modelling

N. Ratkovich a,b,*, W. Horn c, F.P. Helmus c, S. Rosenberger c, W. Naessens d, I. Nopens d,

T.R. Bentzen a
a
Aalborg University, Department of Civil Engineering, Sohngaardsholmsvej 57, DK-9000 Aalborg, Denmark
b
University of Los Andes, Department of Chemical Engineering, Cra 1 No. 18A-12, Bogota, Colombia
c
Osnabrück University of Applied Sciences, Faculty of Engineering and Computer Science, P.O. Box 1940, 49009 Osnabrück, Germany
d
Ghent University, BIOMATH, Department of Mathematical Modelling, Statistics and Bioinformatics, Coupure Links 653, B-9000 Ghent,
Belgium

article info abstract

Article history: Rheological behaviour is an important fluid property that severely impacts its flow
Received 17 July 2012 behaviour and many aspects related to this. In the case of activated sludge, the apparent
Received in revised form viscosity has an influence on e.g. pumping, hydrodynamics, mass transfer rates, sludgee
8 November 2012 water separation (settling and filtration). It therefore is an important property related to
Accepted 13 November 2012 process performance, including process economics. To account for this, rheological
Available online 23 November 2012 behaviour is being included in process design, necessitating its measurement. However,
measurements and corresponding protocols in literature are quite diverse, leading to
Keywords: varying results and conclusions. In this paper, a vast amount of papers are critically
Rheology reviewed with respect to this and important flaws are highlighted with respect to
Activated sludge rheometer choice, rheometer settings and measurement protocol. The obtained rheograms
Apparent viscosity from experimental efforts have frequently been used to build viscosity models. However,
Rheometer this is not that straightforward and a lot of errors can be detected with respect to good
Measurement protocol modelling practice, including fair model selection criteria, qualitative parameter estima-
Good modelling practice tions and proper model validation. These important steps are however recurrently
Membrane bioreactor violated, severely affecting the model reliability and predictive power. This is illustrated
with several examples. In conclusion, dedicated research is required to improve the
rheological measurements and the models derived from them. At this moment, there is no
guidance with respect to proper rheological measurements. Moreover, the rheological
models are not very trustworthy and remain very “black box”. More insight in the physical
background needs to be gained. A model-based approach with dedicated experimental data
collection is the key to address this.
ª 2012 Elsevier Ltd. All rights reserved.

* Corresponding author. Aalborg University, Department of Civil Engineering, Sohngaardsholmsvej 57, DK-9000 Aalborg, Denmark.
Tel.: þ45 9940 8572.
E-mail address: nr@civil.aau.dk (N. Ratkovich).
0043-1354/$ e see front matter ª 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.watres.2012.11.021
464 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2

Contents

1. Introduction and incentives to measure and model rheology for activated sludge applications . . . . . . . . . . . . . . . . . . . . 464
2. Rheology of activated sludge: state-of-the-art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 465
3. Critical considerations with regard to rheological experiments for AS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473
3.1. Nature of activated sludge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473
3.2. Rheological measurement devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473
4. Critical considerations of good modelling practice when modelling AS rheograms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 475
4.1. Goal definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
4.2. Model selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
4.3. Parameter estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
4.4. Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479
5. Conclusion and perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480

1. Introduction and incentives to measure
and model rheology for activated sludge
applications
Rheological measurements are state-of-the-art in a multitude of
engineering disciplines. For homogenous, non-dispersive fluids,
viscosity represents a fluid property which can be measured and
transferred into plant design. Indeed, designs of equipment
(pumps, mixers, aeration systems, etc.) applied in conjunction
with these liquids are mostly based on their thermo-physical
properties. These fluids can either behave as Newtonian
liquids, i.e. constant proportionality between shear rate and
shear stress, or non-Newtonian if this is not the case. However,
in wastewater treatment, the rheological behaviour of activated
sludge (AS) is much more complex as it is composed of water and
dissolved wastewater constituents as the continuous phase and
sludge flocs, particulate wastewater constituents and biological
products (i.e. exocellular polymeric substances (EPS)) as the
dispersed phase. Despite this complexity, the rheological
behaviour of AS is an important property, which is backed up by
numerous studies performed on the topic (see further). Indeed,
viscosity interferes with (1) sludge pumping (i.e. recycle flows),
(2) bioreactor hydrodynamics (i.e. mixing), (3) oxygen transfer,
(4) secondary settler hydrodynamics, (5) membrane filtration
and (6) sludge dewatering. Sludge rheology is thus crucial for AS
management in wastewater treatment plants, especially in
transportation for the calculation of pressure losses in pipes and
pump selection (Tchobanoglous et al. 2003), and for the design of
aeration systems (Cornel et al. 2003; Seyssiecq et al. 2003).
Regarding sludge sedimentation, hydrodynamics in secondary
settlers, especially in the sludge blanket where high solids
concentrations prevail, are crucial for their performance (De
Clercq, 2003; Schumacher, 2006; Brannock et al. 2010a). Sludge
rheology hence interferes heavily with treatment performance
and operational costs as well as with the system design (i.e.
dimensioning of pumps and blowers). Besides the sludge
management inside the wastewater treatment chain, rheology
is also important for further sludge handling like dewatering or
biogas production from sewage sludge. To increase biogas
production, it is necessary to recycle and recirculate digested
sludge in order to mix it with incoming sludge. The flow
rate in the recirculation circuits has to be very large and
rheology is needed to calculate head losses and pumping power
(Slatter, 2001).
Pressure drop of pipe flow for a Newtonian liquid like water
is straightforwardly obtained from the pipe diameter and the
flow velocity as well as the friction factor depending on the
flow regime (laminar or turbulent). Under laminar flow
conditions, the friction factor of the Newtonian liquid is
inversely proportional to the Reynolds number (Re), and the
pressure drop is proportional to the product of the velocity
and viscosity. As the Reynolds number increases beyond
a critical value (Re z 2100), the flow becomes turbulent and
results in smaller friction factors (Bird et al. 2001). In contrast,
non-Newtonian liquids usually reach turbulent conditions at
much higher fluid velocities due to their elasticity. Proff and
Lohmann (1997) proposed pipe friction factors and pressure
losses of AS that were characterized by sludge thinning
properties (power law approach). Tchobanoglous et al. (2003)
illustrated the use of Hedstrom number (He) and the Rey-
nolds number to determine the impact on the friction factor
and how it affect the pressure drop for AS with Bingham
properties (Fig. 1). This presents a critical problem as settle-
able solids accumulate at the bottom of horizontal pipe
sections and eventually lead to pipe blockage (Slatter, 2004).
Settling also might pose a problem within the biological
reactors, where sufficient mixing is crucial in order to achieve
good biological treatment rates. Settling and dead zones due
to elevated effective viscosities will lead to a loss of bioreactor
performance. Hence, the choice of mixing equipment and
propellers during design is important, affecting energy
requirements. Optimizing design and operation thus require
good knowledge with respect to AS rheology.
Within secondary settlers, on the other hand, where sedi-
mentation is required, hindered and compressive settling will
lead to large concentration gradients in the sludge blanket,
which might lead to very different rheological behaviour at
different locations in the blanket. A good understanding of
sludge rheology of high concentration sludge at low shear
rates is crucial in this case. This occurs at the bottom of the
sludge blanket where the sludge needs to move towards the
sludge hopper governed by the bottom slope and/or the
scraper. Hence, the design of the sludge removal structures is
very dependent on the rheological behaviour of the sludge.
 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2
Fig. 1 e Friction factor for AS analysed as a Bingham plastic (see Eq. (4)) in function of the velocity (u) and the tube diameter
(d ). The dimensionless numbers are defined by He[rd2 so =k and Re[rud=k (Tchobanoglous et al., 2003).
The higher the mixed liquor suspended solid (MLSS)
concentration of the AS, the more significant rheological prop-
erties and their impact on energy consumption and process
control become. This is most important for wastewater treat-
ment in membrane bioreactors (MBR) with MLSS concentra-
tions around 10e15 g/L, in comparison to 3e5 g/L in
conventional wastewater treatment plants (WWTP). Many
authors state the increase in sludge viscosity with increasing
MLSS concentrations (Guibaud et al. 2004; Laera et al. 2007;
Pollice et al. 2007; Eshtiaghi et al. 2012). Furthermore,
membrane filtration of mixed liquor requires a cross flow along
the membrane surface for fouling control. The cross flow is
either applied by pumping or by the airlift principle. Here, the
interaction between bubble formation, flow field and shear at
the membrane surface requires detailed studies on AS viscosity.
Fabiyi and Novak (2008) investigated the effects of elevated
solid concentrations on the mass transfer characteristics for
a wide range of mixing and aeration equipment. They observed
a significant dependence of bubble size on viscosity values
(changed through solids concentration). In a low viscosity
system (Fig. 2a), they observed that micro-bubbles were
formed, and the bubble plume diameter was similar to the
sparger diameter. As the solids concentration was increased
(Fig. 2b), a steady decline in the diameter of the bubble plume
was observed (the bubble plume diameter stabilized at about ¼
of the aerator diameter), as well as an increase in the bubble
diameter (MLSS > 12 g/L). The bubbles appeared to have coa-
lesced to large interconnected bubbles contained within the
plume. The authors suggest that bubble coalescence occurs due
to the effect of the viscosity of the liquid continuous phase on
the critical detachment bubble diameter. Obviously, the
difference in plume diameter and bubble size will have an
important impact on the oxygen transfer efficiency due to
a smaller volume of liquid exposed to bubbles and a smaller
surface area of gas bubbles. In addition, there will be an effect of
contact time between liquid and gas bubbles and the influence
of the turbulence within the bubble column. Nevertheless, to
quantify this effect, more knowledge of AS rheology is required.
465
Due to the observed significant impact of rheology on
different processes, quantifying (sludge) rheological behav-
iour to embed in process models also surfaced as being
important in order to trust model predictions: e.g. computa-
tional fluid dynamics (CFD) modelling of sludge processes,
sludge pumping models, filtration models, models describing
the impact of rheology on oxygen transfer efficiency, settling,
mixing, flocculation, etc. (Germain et al. 2007; Barbot et al.
2010; Brannock et al. 2010b; Wang and Dentel, 2010). For
simplicity, sludge viscosity is often replaced by the viscosity of
plain water or in some cases treated as non-Newtonian
including the impact of MLSS in calculations. It should be
pointed out that this is a rough estimation and might cause
intolerable deviations from realistic process behaviour. The
latter implies that the predictive power of models is highly
affected by this and, hence, any decision based on it.
In summary, it can be stated that knowledge of AS rheo-
logical behaviour is crucial both in optimizing design and
operation of sewage treatment plants. These incentives have
led to a significant amount of research on the topic in litera-
ture. The objective of this review paper is to provide a critical
overview with regard to experimental investigation of sludge
rheology as well as mathematical modelling. The first section
will provide an overview of the state-of-the-art of measuring
and modelling AS rheology. This is followed by a section
highlighting potential issues related to measuring AS rheology
and a section on typical flaws when modelling AS rheology.
The paper concludes with perspectives for further research in
both the areas of measuring and modelling AS rheology.
2. Rheology of activated sludge: state-of-the-
art
Rheology is the science of the flow phenomena through the
viscous characteristics of a specific liquid. Dynamic viscosity
(m, Pa s), often simply referred to as viscosity, is a measure of
the flowability of a fluid. It is used to study the fundamental
466 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2

Fig. 2 e (a) Bubble characteristics in a low viscosity system (water). The width of the bubble plume is approximately equal to
the diameter of the sparger (Ds), and (b) Bubble characteristics in a high viscosity system (Carboxymethyl cellulose, CMC).
The width of the bubble plume is significantly smaller than the diameter of the sparger and the mean bubble size is much
larger (Fabiyi and Novak, 2008).

properties of suspensions such as the size and shape of
particles, degree of hydration, state of aggregation and attri-
tion, rigidity of particles, and forces acting between particles
(Dick and Ewing, 1967). The greater the viscosity, the more
viscous and less flowable is the fluid, which means that the
molecules in higher viscous liquids are more strongly bound
to each other and thus less freely moveable. Viscosity is
a property that influences the hydraulic regime and transport
phenomena. It is defined by the ratio between shear stress
_ s1), defined by Eq. (1).
(s, Pa) and shear rate (g,
s (depending on the dispersing phase flow characteristics,
m¼
g_
Furthermore, the interaction between molecules is affected
by temperature. Therefore, temperature dependent relation-
ships for viscosity have been defined. E.g. for water, this is
given by an exponential Arrhenius type relation defined in Eq.
(2) (Linstrom and Mallard, 2009; Yang et al. 2009).


16440:4488
Rgas T
mwater ¼ 1:2182$106 e
where Rgas is the universal gas constant (¼8.3145 J K1 mol1)
and T is the temperature (K). Eq. (2) is valid in the range of
0e100  C (273e373 K).
A plot of shear stress versus shear rate, obtained through
a rheological measurement (which imposes either shear rate
or shear stress and measures shear stress or shear rate), is
called a rheogram and some typically observed relationships
are shown in Fig. 3. For Newtonian liquids, e.g. water, a linear
proportionality prevails, resulting in a constant viscosity over
the entire shear rate range. For non-Newtonian liquids,
viscosity depends on the applied shear rate. Therefore, this
viscosity is referred to as apparent viscosity. Particulate
suspensions, such as AS, exhibit complex flow behaviour
(1) dispersed phase effects and particleeparticle interactions)
and behave like non-Newtonian fluids. As a consequence, the
simple Newtonian relationship (Eq. (1)) is unfortunately not
valid for AS.
In literature, numerous rheological models exist for
a multitude of research domains, all predicting the apparent
viscosity with different degrees of complexity (i.e. the number
of parameters they contain). Most of them are pure empirical
models (correlation models) which either describe shear
(2)
stress or apparent viscosity as a function of its predefined
predictors (shear rate, yield stress). They are mainly inspired
by the shape of the rheogram obtained experimentally, which
can be described mathematically by a power law function or
 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2 467

Table 1 e Different widely used non-Newtonian shear

stresserate relationships or explicit apparent
viscosityeshear rate relationships. s0 is the yield stress
(Pa), n the flow behaviour index (e), k the flow consistency
index (Pa sn), mN the infinite rate apparent viscosity (Pa s),
m0 the zero shear apparent viscosity (Pa s), l the (Cross)
time constant (s) and m the Cross rate constant (e).
Model Equation

Power-law (Ostwald de Vaele) s ¼ kg_ n

(3)
Shear-thinning
(pseudoplastic) (n < 1)
Shear-thickening (dilatant)
(n > 1)
Bingham s ¼ s0 þ kg_ (4)

Herschel and Bulkley s ¼ s0 þ kg_ n (5)

(0 < n < N)
Casson N g
s0:5 ¼ s00:5 þ m0:5 _ 0:5 (6)

Sisko m ¼ mN þ kg_ n1 (7)

m  mN 1
Fig. 3 e Shear stresseshear rate diagram: 1: Shear-
Cross ¼ (8)
_ m
m0  mN 1 þ ðlgÞ
thickening (dilatant), 2: Newtonian, 3: Shear-thinning
(pseudoplastic), 4: Bingham plastic and 5: Casson plastic  n1
Carreau m  mN 2
fluids. _ 2
¼ 1 þ ðlgÞ (9)
m0  mN

a variation thereof, e.g. including an offset (Table 1). For non-

dispersive liquids, parameters are usually obtained by fitting
the equations to experimentally obtained rheograms.
However, for dispersive liquids, rheograms have been
observed to differ depending on other factors like concentra-
tion of the dispersion.
Hence, the parameters from the models (Table 1) should
not be calibrated in a scalar way, since they are known to be
a function of (at least) temperature and sludge concentration
(defined by MLSS). For this purpose, mostly empirical sub-
models as functions of MLSS and temperature have been
defined for the different parameters in the rheological models.
Alternatively to these general models applicable to numerous
fluids, complete new empirical correlations are derived for the
specific fluid under study, e.g. AS, directly based on the
influencing variables (temperature, MLSS, shear rate).
Examples of these specific empirical apparent viscosity
models and of general apparent viscosity models using empirical
parameter functions can be found in Table 2. Note that specific
models (based on research specific correlations) can end up with
a similar structure as general literature models; e.g. apparent
viscosity increasing exponentially with MLSS and decreasing
exponentially with shear rate, as often observed, would lead to
a specific model m ¼ ða MLSSb Þðc g_ d Þ ¼ ðac MLSSb Þg_ d , which in
fact is equal to the general Ostwald de Vaele model with an
exponential expression for k as a function of MLSS (with d < 1)
(Eq. (3)). It is noteworthy that these models contain quite
a number of parameters that are typically highly correlated.
To calibrate a general or specific model, rheological
measurements are required. Most rheometers use a rotational
system to generate different shear rates. These shear rates are
calculated using the MetznereOtto calibration (g_ ¼ KMO N),
stating that the rotational stirring speed (N ) is proportional to
the shear rate, with KMO the MetznereOtto constant (Metzner
and Otto, 1957). In full-scale AS systems, shear rates are
imposed by advective flow accompanied by mixing devices
(in anaerobic and anoxic zones). However, in aerated zones,
shear rate is also influenced by air bubbles introduced at the
bottom of the bioreactor (in case of fine bubble plate aeration;
not valid for surface aeration). Due to their upward movement
and the friction between the bubbles and the ambient
suspension, they will impact the movement of the suspen-
sion. The effective fluid movement is obviously a complex
three-dimensional one. If the shear is provided by aeration
rather than stirring, urging the need for another relationship,
which is provided by Popovic and Robinson (1984) as
g_ ¼ 5000UGr .The dominating movement can be described as
a shear rate caused by shear stress as described above.
However, the suspension is also exposed to elongation stress.
When a bubble passes, the suspension is accelerated which
results in an extensional stress (s) and creates an extensional
strain rate (_ε). In analogy to shear apparent viscosity, exten-
sional viscosity (hE ) is defined by Eq. (17).
s ¼ hE ε_ (17)
Extensional strain occurs in many flow situations such as
flow through orifices. Considerable efforts have been made to
investigate extensional viscosity in the past (Oliver and Bragg,
1974; Usui and Sano, 1981; Jones and Rees, 1982).
Due to the relevance of sludge rheology on process design
and performance, many studies on AS apparent viscosity are
published in literature. Table 3 gives an overview on results of
42 publications. It shows the type of AS, the rheometer used
468 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2
Table 2 e Different forms of correlation relationships between combinations of shear rate, MLSS and temperature with
apparent viscosity. In these equations, a, b, c and d are parameters to calibrate and Ea is the activation energy (kJ molL1).
Model Equation
_ MLSSÞ
m ¼ f ðg;
eða MLSS Þ g_ ðc MLSS Þ
b d
a MLSSb g_ c
MLSSb
a (12)
g_
MLSS
a þ b MLSS þ c (13)
g_
m ¼ f ðMLSS; TÞ a MLSSb Tc
Ea
a MLSSb eRgas T
_ MLSS; TÞ
m ¼ f ðg; MLSSb Rgas
Ea
a e T
g_
on the experimental measurements (brand and geometry),
MLSS range, shear rate range, temperature range, the model
used based on Table 1, the most important results from the
research and if the measurement protocol is included on the
reference. This measurement protocol was divided into three
stages: (i) not included, (ii) partially included and (iii)
completely included on the reference. A complete protocol is
defined as a complete experimental procedure where: (i) shear
rate or shear stress controlled measurement procedure, (ii)
upward and/or downward ramp, (iii) distribution of reading
steps and duration of each step, (iv) reading range, and (v)
temperature are mentioned. Also, the samples can be
measured on site, in the laboratory (after transport), undiluted
or diluted. Given the fragile nature of biological AS flocs, all of
these parameters can potentially influence obtained rheo-
grams and conclusions based on them.
In all studies, the viscous characteristics of AS are identi-
fied as non-Newtonian, more specifically shear-thinning (see
Table 3). A great number of authors identified a yield stress
(Mikkelsen, 2001; Christopher, 2002; Laera et al. 2007;
Brannock et al. 2010b) while other authors did not detect
a yield stress (Moeller and Torres, 1997; Wang and Dentel,
2010). Yang et al. (2009) report the shear-thinning behaviour
of AS in conjunction with a static and dynamic yield stress for
MBR and conventional activated sludge (CAS) systems. The
transition from solid-like to fluid-like behaviour is interpreted
by a critical stress below which the material behaves as an
elastic solid. The change in this rheological behaviour is
explained by first an alignment and then a rupture of the flocs.
Furthermore, thixotropic properties, i.e. different apparent
viscosity behaviour when increasing shear rate compared to
decreasing shear rate, are reported too (Guibaud et al. 2004).
MLSS and temperature contribute significantly to the
apparent viscosity (Yang et al. 2009; Garakani et al. 2011) while
rheological properties among different AS from different
water treatment plants clearly show heterogeneity from
References
Rosenberger et al. (2002); Laera et al. (2007)
(10)
(11) Yang et al. (2009)
Garakani et al. (2011)
Laera et al. (2007)
(14) Yang et al. (2009)
(15) Yang et al. (2009); Garakani et al. (2011)
Garakani et al. (2011)
(16)
system to system (Spinosa and Lotito, 2003; Yang et al. 2009).
The viscous characteristic is usually modelled using either the
pseudoplastic power-law or Bingham plastic rheological
models (see Table 3). Hence, the general rheological behaviour
of AS is not really an issue, but the exact apparent viscosity
value, which is typically the most critical, under different
circumstances and in different systems shows a broad vari-
ability in the literature. Indeed, this is problematic when the
exact value is required for design and operation purposes. The
influence of aeration on the rheological behaviour (elonga-
tional rheology) is not a thoroughly studied domain. As an
exception on this, Seyssiecq et al. (2008) studied the influence
of aerational shear on top of the traditional rheometer
shearing and found that aeration provided a significant
decrease in apparent viscosity measurements in the low shear
rate range. However, this decrease was rather independent of
aeration intensity. For high shear rates, no impact of aeration
was concluded for the entire range of aeration intensities
studied.
As a reason for this bothersome heterogeneity, Campbell
and Crescuolo (1982) highlighted that rheograms are very
dependent upon the rheometer geometry, operation and the
sample conditioning. Also, they pointed out that comparing
rheological data is extremely risky unless detailed informa-
tion on the procedure is available. As well, they mentioned
that (de)flocculation of AS particles in the rheometer is
a function of the conditioning, shear rate and shear time.
Therefore, increasing any of these variables will tend to
increase the degree of particle breakdown. Moreover, the time
during which an AS sample is subjected to shear, appears to
be equally as important as the shear rate. Seyssiecq et al.
(2003) made a concise literature review devoted to the rheo-
logical characterization of AS. They presented the rheological
equations used by the different authors to represent the
shear-thinning, viscoplastic or viscoelastic properties of AS.
Next to that, the thixotropic behaviour that frequently
Table 3 e Overview on results of 42 publications regarding AS rheology.
Reference Type of AS Brand Geometry Measurement Solid Shear Temp. Model Results
protocol concentration rate ( C)
provided? (g/L) (s1)

Sanin (2002) WWTP Brookfield LVDVII Sp Partial 2e18 1.8e73.4 25 PL  Solution pH affects AV
of AS.
 Lowest AV is observed at lowest
pH and vice versa.
 AV is found to decrease with
increasing conductivity of AS.
 AV is found to decrease with
removal of polymers by centrifu
gation (EPS affects AV)
Rosenberger et al. MBR Haake VT 550 DCC Partial 2.7e47 0e2200 21 PL  AV function of MLSS
(2002)
Yen et al. (2002) Dewatering Brookfield DV-III Sp Partial 0.27e0.64 0e120 NM NM  AV is related to the networks

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2
strength
Spinosa and Lotito Synthetic Kasumeter TB Partial 3.51e44.67 NM 20 NM  Yield stress of AS varies
(2003) suspensions of HaakeRheotest CC exponentially with MLSS.
kaolin and quartz RV 2.1
sand in water
Tixier et al. (2003a); WWTP PAAR Physica CC Partial 3.1e6.3 0e800 20 PL & B  AV affected by MLSS
Tixier et al. (2003b); system MC100  Settleability properties are related
Guibaud et al. (2005) to AV (changes in floc structure).
 Reduced hysteresis area (rHa)
was found to vary greatly according
to nature of the AS (i.e.
filamentous AS)
Guibaud et al. (2004) WWTP PAAR Physica CC Partial 2.9e11.3 0e500 20 B  AS presents very weak thixotropic
system MC100 behaviour.
Chen et al. (2005) WWTP HaakeRheostress CP Partial NM NM 20 NM  Stressesweep experiments show
RS 75 that corresponding complex shear
modulus is significantly affected
with addition of polymer
(rigid structures).
Pevere et al. (2006) Anaerobic AS PAAR CC Partial 8.3e22.6 200e1000 20 NM  Granular composition and surface
Physicasystem charge interactions between
MC100 granules affect AV.
Mori et al. (2006) WWTP TA instruments CC & DCC Partial 27e57 0e3000 20 PL, B & HB  Dynamic yield stress functions
AR 550 of MLSS.
Laera et al. (2007) Lab-scale MBR Rheotest 2.1, CC Partial 3.7e22.9 3e1300 20 PL & B  B model provided slightly better
HaakenMedingen results than PL model in terms
GMBH of root mean squared error.
Banu (2007) Dewatering Koei Industry Sp Partial 2.5e10 NM NM NM  Rheology can be used to
Floccky Tester control dewatering.
Pevere et al. (2007) Anaerobic AS of PAAR Physica CC Partial 0e23 0e800 22 PL, B & C  Statistical analysis of 2nd
MBR and CSTR system MC100 order polynomial relationship

469
(continued on next page)
 470
Table 3 e (continued )
Reference Type of AS Brand Geometry Measurement Solid Shear Temp. Model Results
protocol concentration rate ( C)
provided? (g/L) (s1)

between rheological parameters

and MLSS found B model more
suitable to describe evolution of
anaerobic AS rheology.
Wu et al. (2007) Lab-scale MBR Model NDJ, TJ NM Partial 4.2e25 940 24 NM  High MLSS also reduced kLa
Environmental in MBR.
Facility Corporation,
Shanghai, China
Li et al. (2008) Lab-scale MBR ThermoHaake NM Partial 3e8 100 24 NM  EPS has a strong impact on
RheoStress1 membrane permeability but no
apparent relationship was found
with AV of AS.

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2
Seyssiecq et al. Lab-scale MBR Contraves Sp Partial 10e35 0.1e100 20 NM  Under low mechanical stirring,
(2008) Rheomat 30 and TA flocs configuration is driven by
instruments AR550 shear rate induced by air plume.
An increase of air flow rate
induces an increase of shear due
to bubbles injection, leading to a
decrease in AV (shear thinning
medium).
 Under high mechanical stirring,
mechanical shear rate prevails
compared to air plume, in that
particular case, if there is
sparging or not, the AV is
constant for a
given mechanical shear rate.
Yang et al. (2009) Lab-scale MBR TA instruments AR NM Partial 2.74e16 25e1000 20 Different models  Found shear-thinning behaviour
2000 (See Table 2) and static and dynamic yield
stresses for MBR and CAS
systems.
 Ca model provides the best
model over the entire MLSS
concentration range, followed
by the HB model for a shear rate
above 25 s1. The B model was
suitable for CAS systems under
low MLSS concentrations (<10 g/L).
Moreau et al. (2009) Full-scale MBR NM TB Partial 3.5e16.4 1000 9.7e27.4 NM  Main factor influencing AV of AS
was MLSS content.
 Temperature did not directly
affect AV of AS.
Wang and Dentel Anaerobic AS Haake RV20 CC Partial 28.43 0e100 25 PL  No yield stress
(2010)
Barbot et al. (2010) Lab-scale MBR Contraves Sp Partial 15 0.1e100 20 PL  AS exhibits a non-Newtonian
Rheomat 30 shear thinning behaviour with and
without sodium azide injection.
 Reaction with sodium azide
induces a decrease of the AV,
due to a deflocculation of the
biological suspension.
Aranowski et al. Anaerobic WWTP Broekfield LVDV IIþ Sp Partial 5 0e100 25 HB  Process of fermentation,
(2010) decrease in content of organic
matter even by 60%,
pseudoplasticity and
consistence ratio increases.
Marinetti et al. (2010) Dewatering HaakeRotovisco Sp Partial 24.23 0e100 NM NM  Rheology of AS has greater
RV20 importance in centrifuge
dewatering which apply
considerable shear forces.
Baudez et al. (2011) Anaerobic digestion RheometricScientific CC Partial 18.5e49 0e1000 25 KelvineVoigt  Digested AS is a shear thinning

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2
DSR200 model and HB & B yield stress fluid, presenting
flow instabilities at low shear
rates, manifesting as shear
banding.
Hammadi et al. Sewage AS WWTP Thermo-Fischer CP Partial 23e28.5 0e2000 5e30 HB  Increasing temperature, both
(2011) MARS II yield stress and flow index
increased while consistency
index decreased.
De Clercq (2003) Sewage AS WWTP Bohlin CVO Rotational Partial 1.24e16 0e500 NM HB with  Impact of floc disruption
stress Papanastasiou’s observed
controlled Adaption  No yield stress found
 Model selection applied to select
best submodel with least
parameters
Moeller and Torres Anaerobic digestion Brookfield LVF Sp NM 7.1e13.7 2e30 NM PL  No yields stress was found.
(1997)  Different AS (origin) shows
different rheological properties.
Paulo Santos (1997) Anaerobic digestion PhysicaRheolab CC NM NM 0.1e1000 32e34 B & HB  B and HB present best fit
CM100 torheological properties.
Slatter (1997; WWTP HaakeRotovisco CC NM 31.7e67.2 100e1000 35 B  AV parameters depend on
Eshtiaghi et al. (2012) MV1P MLSS concentration
Defrance et al. (2000) MBR HaakeRotovisco CC NM 10e12 10e600 20 PL  MBR AS is less viscous than
RV 100 AS from aeration tank
Mikkelsen (2001) WWTP Brookfield LVTDT-II NM NM 3.5e4 36.7e73.4 B  Strong correlations between
dispersed mass concentration,
resistance to filtering (capillary
suction time) and rheological
parameters.
 Resistance to flow is caused by
development of inter-particle
networks.

471
(continued on next page)
 472
Table 3 e (continued )
Reference Type of AS Brand Geometry Measurement Solid Shear Temp. Model Results
protocol concentration rate ( C)
provided? (g/L) (s1)

Slatter (2001); Slatter Thickened AS NM TB NM 35 5e75 NM B  High concentration AS pipelines

(2004) will operate in laminar flow
regime.
Christopher (2002) Anaerobic digester Fann Instrument CC NM 10e50 NM 20 NM  AS samples exhibited yield
Model 35A stresses which had a power
relationship with MLSS.
Hasar et al. (2004) Lab-scale MBR NDJ-Model Sp NM 2.9e12.3 NM 15e30 PL  Shear stresses decreased with
an increase in temperature.
 Temperature and solid
concentration of AS were the
most important parameters that
influence the flow behaviour.

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2
Jin et al. (2006) Municipal WWTP Brookfield LV- Sp NM 2e10 20e200 4 _ b
s ¼ a$lnðgÞ  Impacts of operating conditions
DVIII and liquid properties on the
hydrodynamics and kLa in AS
air-lift reactors.
 Liquid circulation velocity, gas
holdup and kLa decreased as AS
MLSS increased, since the AV
increased.
Pollice et al. (2006); MBR HaakeRheotest 2.1 CC NM 3e30 3e1300 20 PL, B & HB  AV increases less than
Pollice et al. (2007); proportionally for increasing
Pollice et al. (2008) MLSS.
Meng et al. (2007) MBR NM NM NM 2e20 3500 NM NM  MLSS concentration and AV
of AS were two major factors
affecting membrane fouling.
Van Kaam et al. MBR Malvern CP NM 8.5 10e1000 23 NM  AS had a different structure
(2008) Instruments depending on the applied shear
Bohlin C-VOR 200 stress.
 AS was characterized by its
viscoplastic property, which
leads to a possible restructuring
ability when a shear stress lower
than the yield stress is applied.
Kornboonraksa and Lab-scale MBR NM NM NM 9.3 36.1 NM 23e25 m ¼ a MLSSb  AV is a sub-factor which affects
Lee (2009) membrane filtration process
through its good correlation with
MLSS
Zhao et al. (2009) MBR NM NM NM 6e16 NM NM PL  AV function of MLSS
Xia et al. (2009) MBR NM NM NM 3.3e28.6 850e1850 25 PL, B, HB & S  AS showed both shear thinning
and viscoplastic behaviour, and
under various shear rates
different rheological models
 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2 473

appears in concentrated flocculated suspensions was dis-

CC: Concentric cylinder; CP: Cone and Plate; DCC: Double gap concentric cylinders; Sp: Splinder; PL: Power-law; B: Bingham; HB: Herschel and Bulkley; C: Casson; S: Sisko; Ca: Carreau; Cr: Cross; AV:
C model provides the best prediction

local shear rate while the remaining

localizes in a small region with high
cussed. They also tried to link rheological properties with

AV values ranged between 3.6,103

tests showed that the AS rheology
full-scale MBRs, where a modified
PL model was incorporated in the

AV values are strongly correlated

part of fluid behaves like a solid.

physico-chemistry of the flocs, working conditions of
could be chosen to characterize

Comparison between the tracer

Developed a CFD model of two

processes (in which sludges are operated) and the kLa in

At low shear rates, the shear

of AV over the entire MLSS
bioreactors. All these works have clearly demonstrated the
difficulties in performing reliable rheological measurements
had minimal effect. ()
their flow behaviour.

concentration range.
with MLSS content.
on AS. This fact is notably due to the great variability of sludge

and 3.4,102 Pa s.
samples but also to the rheometers and methods themselves.
CFD model.

3. Critical considerations with regard to

rheological experiments for AS



s ¼ s0 ð1  emg_ Þ þ kg_ n

3.1. Nature of activated sludge

Papanastasiou’s

PL, B, HB, C, S,
Adaption
HB with

Ca& Cr

Before discussing AS rheology, the nature of AS needs to be

NM

briefly highlighted. AS is an extremely complex multiphase

fluid containing mainly water (98e99.7 %w/w), solutes (e.g.
carbon, nitrate and phosphate) and suspended flocs with
primarily deformable surfaces (e.g. bacteria, proto- and met-
15e25

azoa, flocs, EPS and tiny gas bubbles). These flocs (see Fig. 4)
NM
25

can undergo continuous reorganization through flocculation

and deflocculation depending on the shear to which they are
0.01e1000

0e1000

exposed (Nopens et al. 2002; Biggs et al. 2003). The higher the
NM

MLSS, the more difficult it is to distinguish between single

flocs. In MBR with MLSS concentrations as high as 10e15 g/L,
these flocs often reveal a loose structure with many cross-
2.8e17.2

2.74e31

links between single flocs.

8

The microscopic structure of AS is subject to constant

changes due to fluctuations in ambient conditions like
temperature, substrates, loading rates, and shear forces
(advective flow, mixing, aeration). All these changes affect the
rheological behaviour and will also make it dynamic over
NM

NM

NM

time. Therefore, sampling conditions, storage times and

measurement procedures are likely to influence sludge prop-
erties and thus the obtained rheograms.

3.2. Rheological measurement devices

NM
CC

CC

Rheological measurements to obtain a rheogram are carried

out in a rheometer, which establishes a relationship between
HaakeRheostress

shear stress and shear rate. There are generally two types of
UDS-200 Anton

rheometers applied in the measurement of AS apparent

viscosity: (i) the rotational and (ii) the tubular (capillary)
rheometer. Each of these rheometers has its advantages and
RS 75

Apparent viscosity; NM: protocol not mentioned.

Paar

NM

drawbacks, which can be found in literature (Slatter, 1997;

Mori et al. 2006). The most important features are presented in
Table 4.
Full-scale MBR

Full-scale MBR

Measurement devices used in rotational rheometers for

sludge analysis are mostly single or double gap cylinders
(bobs) or plate and cone geometries. For apparent viscosity
MBR

measurements in rotational viscometers with single or double

gap cylinders, the following assumptions apply: laminar flow
without vortex formation, no slip, isothermal conditions and
homogeneous fluid (no settling). Gap sizes should be related to
Brannock et al.

Gil et al. (2011)

Garakani et al.

floc size. However, as seen in Fig. 4b, it can be extremely

difficult to define the floc size, as based on the nature of AS the
(2010b)

flocs can be cross-linked among them. When measuring floc

(2011)

size distribution in particle analysers, weak bonds between

flocs are likely broken by the dilution and handling of the
474 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2
Fig. 4 e Microscopic pictures of AS a) from a conventional
AS plant witha MLSS of 3.9 g/L and b) from a pilot scale
membrane bioreactor with a MLSS of 12.2 g/L;
magnification: 403 (Picture: Osnabrück University of
Applied Sciences).
sample. On the other hand, gap sizes have to be as small as
possible in order to avoid settling of particles. The latter is also
affected by MLSS concentration, where higher MLSS leads to
lower settling velocities. Moreover, smaller gap sizes will
induce wall effects that once again reduce settling velocities.
Consequently, the settling problem needs to be treated with
special care when measuring at low MLSS (<2 g/L).
It is also important to consider the flow conditions in the
rheometer in order to avoid measurement errors caused by
vortices and turbulence. The flow in rotational viscometers
can be described in analogy to Couette flow using the
dimensionless Taylor number defined by Eq. (18).
rffiffiffiffiffi
r$Ri $u$s s
Ta ¼ $ (18)
m Ri
where Ri is the inner radius (m) of the bob, u is the angular
velocity (rad s1) and s is the gap between the rotating and the
static cylinder. It is important to highlight that the apparent
viscosity is actually required to compute this number. For
non-Newtonian liquids that obey the power-law, the Taylor
number is defined by Eq. (19).
rffiffiffiffiffi
r$Ri $u$s s
Ta ¼
n1 $ (19)
Ri $u Ri
k
s
Taylor vortices occur for Taylor numbers larger than the
critical Taylor number (Tac  41.3). The flow field only
becomes fully turbulent for Ta  105. The formation of Taylor
vortices leads to a sudden increase in apparent viscosity
which may be erroneously attributed to shear-thickening
behaviour (Rios et al. 2007). This is clearly illustrated in
Fig. 5. As can be observed, the onset of turbulence kicks in at
higher shear rates for elevated MLSS.
From Eq. (19), it is observed that the flow behaviour and
flow consistency index need to be known a priori to determine
the Taylor number. For that, the only option is to perform
rheometrical measurements, and find out when the apparent
viscosity starts increasing (as shown in Fig. 5) and analyse the
data until that point (laminar regime). In addition, the gap size
will affect the onset of turbulence; therefore a small gap size is
recommended, but still larger than the biggest floc particle.
Seyssiecq et al. (2003) and Mori et al. (2006) showed
a detailed investigation on how to perform a correct rheolog-
ical characterization of a complex material such as the AS
from a WWTP. They highlighted the importance of the defi-
nition of relevant measurement geometries and experimental
procedures, which will be referred to as measurement
protocol. They used two different measurement geometries,
i.e. concentric cylinders (CC) and double gap (DG) concentric
cylinders. They found that the CC system is suitable for the
characterization of AS. However, the gaps of the DG are too
narrow, leading to a blockage of the system.
The significance and comparability of published rheological
data therefore strongly depends on the interaction between AS
nature on one hand and rheometer type, measurement device,
and experimental procedures (i.e. the measurement protocol)
on the other hand. Only very few authors specify their
measurement protocol sufficiently, as can be seen from Table 3.
As a consequence, the comparison of rheograms obtained by
different authors is nearly impossible.
The impact of measurement protocol on the obtained
rheograms will be demonstrated here by three cases: (i) In Fig. 6,
a comparison of two different rheometers both equipped with
concentric cylinders as measurement geometries is shown.
The difference between these two rheometers is the gap and
the diameters of the bob/cup. From Fig. 6, the difference
between the two rheometers is obvious in spite of the fact that
the AS samples were identical (the measurement on the two
rheometers were performed in parallel). It is important to
highlight as well that the onset of Taylor vortices happens at
different shear rates for different rheometers as described in
Fig. 5. (ii) In Fig. 7, the difference between a shear-rate and
a shear-stress controlled measurement with the same rheom-
eter is illustrated. Not only are obtained apparent viscosity
readings different for identical shear rates, also the quality of
the curves is different: in the flow curve for the shear rate
controlled measurement, both a static and a dynamic yield
stress can be clearly detected. For the shear stress controlled
 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2
Table 4 e Advantages and drawbacks of different types of rheometers (Slatter, 1997; Seyssiecq et al., 2003).
Rheometer Advantages
Rotational  Time dependent effects (e.g. thixotropy) can be measured.
 Widely spread, commercially available devices.
 Small-volume sample required.
 Rheograms can be obtained directly by data transfer to a PC.
 Simple determination of yield stress
 Simple adjustment of shear rate duration of measurement,
intervals between measurements, etc.
Tubular  Mechanically simple (flow in a tube).
 Larger shear stresses-rates can be attained depending on
flow velocity.
 Measurements in laminar and turbulent flow regime.
 Rheograms are obtained from a series of coordinates of
velocity and pressure drop.
 Diameter dependent effects can be measured.
 Low cost
 In-situ testing
measurement, the dynamic yield stress is not observed. (iii)
Fig. 8 shows apparent viscosity curves measured with different
shear rate ramps. Chosen time intervals for each reading were
logarithmic and identical in both measurements. In Fig. 8a the
apparent viscosity was increased from 1e1000 s1 and then
lowered back to 1 s1, while the hysteresis in Fig. 8b was
1e100 s1 and back to 1 s1. It becomes apparent that the
history of experienced shear (time and maximum shear rate)
has an impact on current AS apparent viscosity.
These three examples demonstrate the sensitivity of
experimental results of AS apparent viscosity to equipment
and measurement procedure. When measuring identical
sludge samples either in different rheometers (see case i) or in
the same rheometer but by varying measurement protocols
(see case ii and iii), different results regarding values of
apparent viscosity for certain shear rates, and rheological
Fig. 5 e Taylor vortices development for shear rates beyond
the critical shear rate TA rheometer type AR2000 (gap
1 mm) (Data: Ghent University).
475
Drawbacks
 The annular gap must be larger than the largest
particles on the sample. However, it has to be as
small as possible in order to minimize correction
factors and avoid eddies formation (turbulence).
 There is no accurate indication of the transition
between laminar and turbulent regimes (i.e. Taylor
number).
 Centrifugal forces can cause changes on solids size
distribution (deflocculation) and concentration
gradient in the measuring gap.
 Sedimentation can cause measurement errors
(more problematic at lower AS concentrations)
 Expensive
 Sample is subjected to varying shear stress-rate
over tube cross-section. However, these parameters
are calculated at the wall (capillary tubes).
 Cannot measure time-dependent effects
(i.e. thixotropy)
 Large sample volumes are required.
properties such as yield stress and thixotropy will be obtained.
As a result of these obvious sensitivities, it seems impossible to
compare apparent viscosity data of different research groups
which all use different devices, sludges and protocols. These
are some of the reason among others for the myriad of different
rheological models and parameter values that have been
proposed in literature (Tables 2 and 3). Before even considering
rheological modelling, it is therefore clear that guidelines for
proper rheological data collection are crucial. The question
arises whether a “correct” and unique measurement exists. It
might as well be that different ones are applicable under
different circumstances or for different goals (see further).
4. Critical considerations of good modelling
practice when modelling AS rheograms
Mathematical modelling aims at translating a given system
into a set of mathematical equations that is able to describe
the system’s behaviour with a certain level of detail. It has
proved itself a valuable tool for system analysis (knowledge
build-up) or system optimization in various domains such as
wastewater treatment (Gujer, 2008) and has also been used for
modelling of rheograms as discussed earlier. Furthermore,
modelling can form the basis for a sensitivity analysis (Saltelli
et al., 2000), scenario analysis (Nopens et al., 2007) or uncer-
tainty analysis (Belia et al., 2009; Benedetti et al., 2012;
Cierkens et al., 2012) for the more experienced modeller.
Despite the enormous variety of processes that can be
modelled, and the diversity of models available, the modelling
process is comprised of an almost universal algorithm (Fig. 9)
(Dochain and Vanrolleghem, 2001). First, the goal of the
modelling exercise should be defined in order to build a model
able to answer the formulated research question(s). This is
very important as different goals might lead to different
476 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2
Fig. 6 e Shear rate vs. shear stress of an MBR AS for two
MLSS-concentrations measured with the Bohlin rheometer
type CVO50 (gap 1.25 mm) and with the TA rheometer type
AR2000 (gap 1 mm). (Data: Ghent University).
Fig. 7 e Comparison of shear-rate and shear-stress
controlled measurement protocol with an MBR sludge with
5.8 g/L, a) apparent viscosity curves, b) flow curves;
measured in an Anton Paar rheometer type MCR101 with
double gap cylinder (0.5 mm gap size) and air bearing
(Data: Osnabrück University of Applied Sciences).
Fig. 8 e Apparent viscosity curves of an MBR AS with
5.7 g/L. The same sample was measured a) with a shear
rate ramp of 1e1000e1 and b) 1e100e1 sL1 in an Anton
Paar rheometer type MCR101 with double gap cylinder
(0.5 mm gap size) and air bearing (Data: Osnabrück
University of Applied Sciences).
models. Hence, there is not such a thing as “the best model” to
describe a certain system. Next, the framework has to be set
up (type of equations, system boundaries, level of detail, .) to
build the model structure and a correct numerical solution
needs to be pursued, also a potential source of errors (Bürger
et al., 2011). If several candidate models are available, one
has to select the most appropriate model, using the model
selection techniques at hand (Dochain and Vanrolleghem,
2001). Depending on the selection technique used, the model
parameters need to be estimated a priori or a posteriori. Note
that all steps, except for the goal definition, require knowledge
and/or experimental data, which need to be collected with
care since ‘garbage in leads to garbage out’. Finally, because
the entire modelling process exhibits several dangers and
pitfalls (overparameterization, unrepresentative calibration
data, parameter identifiability issues, low data quality,.), it is
important that the model is properly validated after its
Fig. 9 e Different steps in a general modelling scheme.
 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2 477

construction. This is the final test to ensure that the model is practice perspective, this is not sound at all as the model will
a correct mathematical translation of the system under study. likely over fit the data rather than represent the true under-
If validation fails, one has to analyse the problem and iterate lying relation, rendering it into a mere data fitting exercise
the modelling steps until a validated model is obtained. rather than knowledge build-up. This over fitting is typically
Although the modelling process is rather straightforward, it not reported as parameter estimates are usually not inspected
must be stated that many studies violate one or more of the for their quality and degree of correlation (i.e. identifiability
steps and end up with erroneous conclusions. That is why we problem). An important problem related with these models is
purposely show this and give some examples where rheo- that they have low predictive power and are likely unable to
logical studies have violated this “good modelling practice” pass a validation test. This actually renders them useless.
(GMP). The conclusions based on the latter studies should be Rosenberger et al. (2002) and Petersen et al. (2008) are excep-
handled with care. tions that do report on these issues as discussed below.
In the following paragraphs, current rheology modelling will This danger of over parameterization is clearly illustrated by
be discussed in view of GMP: the appropriate modelling steps Rosenberger et al. (2002), who compared the HerscheleBulkley
will be discussed and pitfalls will be highlighted. Results will be approach (3-parameter model) to the Ostwald equation
compared between the different researches to provide an (2-parameter model) for measurements on AS samples from nine
overview of the available rheology modelling knowledge. Note MBR installations (municipal and industrial). Since rheograms
that Seyssiecq et al. (2003) listed and discussed rheology were measured for different MLSS concentrations, the parame-
modelling research up till 2002, which will not be repeated here. ters of both models were calculated in function of this MLSS
concentration. For the Ostwald model, satisfactory regressions
4.1. Goal definition were found for both k and n (Fig. 10a). However, including a third
parameter s0 , as required in the HerscheleBulkley model,
Roughly, rheology models have been used for three distinct
purposes: (1) Knowledge build-up, where one investigates the
detailed fluid behaviour within the system. Here, it is impor-
tant to identify the dominating characteristics of the fluid’s
apparent viscosity (type of non-Newtonian behaviour, pres-
ence of yield stress, time-dependency) as well as the major
influential variables on these characteristics (e.g. MLSS, tem-
perature,.). Typically, this goal is translated towards finding
correlations and developing regressions, the latter being
exclusively empirical or semi-empirical in nature (see Tables 1
and 2). The ultimate use of such models is studying the
correlations to deduce rheological phenomena; (2) System
design: CFD studies can use a specific rheological submodel to
predict the hydrodynamics of a system more accurately. With
this information at hand, systems can be optimized for their
compactness, costs or efficiency before construction; (3)
operational optimization of existing systems: scenario anal-
ysis can provide the necessary information to optimize energy
usage, yield, product quality, avoidance of dead zones,.
Even though different goals exist, the models currently
applied for modelling sludge rheology are mostly the same.
However, in literature, even for a single model structure,
parameter sets can vary significantly, indicating that more in
depth knowledge is still lacking. In other words, important
influencing processes/variables are not yet incorporated in the
models. On the other hand, it has been demonstrated that
measurements are not performed in a standardized way and that
this can have significant impact on the model calibration as well.

4.2. Model selection

Since a multitude of models is available in literature,

modellers tend to apply multiple approaches and choose
the most appropriate one for further use. Often, they are
guided by performance indicators like R2 or RMSR values,
which systematically point towards the most parameterized Fig. 10 e Parameters k and n of the Ostwald de Vaele
model (e.g. the Cross or Carreau model). This is to be expected, model, estimated as function of MLSS (a) and parameters k,
since more parameters create more degrees of freedom to fit n and s0 of the HerscheleBulkley model as function of MLSS
a function to experimental data. From a good modelling (b) (Rosenberger et al., 2002).
478 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2
introduced serious additional scatter on the estimation of k: the
parameter became practically unidentifiable, meaning that not
enough information is present in the data to find a unique set of
estimated parameters (Fig. 10b) (Dochain et al., 1995). It was
furthermore stated by the authors that values for s0 were signif-
icantly lower than those reported in literature. Therefore, it was
soundly decided to select the Ostwald model for further calcula-
tions. However, many cases in literature exist where this good
modelling practice is severely violated or not incurred. A brief
review is given below.
Laera et al. (2007), reported scatter in s0 for the Her-
scheleBulkley model and stated this to be due to the estima-
tion method, i.e. a least squares logarithmic regression. Since
the other models in the comparison exercise, Ostwald and
Bingham, had only two parameters, the phenomenon of over
parameterization was more likely the cause for this finding.
The same authors put forward an additional problem of model
selection based on R2. As models are allowed to compensate
faults, it should be checked from a visual inspection whether
the residuals are equally distributed along the curve (Fig. 11).
Ideally, remaining residuals should only be due to measure-
ment error and normally distributed with zero mean. If this is
not the case, the model structure is inadequate and fails to
incorporate all important processes to describe the observa-
tions. Hasar et al. (2004) compared a Newtonian model, Bing-
ham model and power law. The power law was indicated as
best model for its high R2 values, while the Bingham and
Newtonian model showed an almost identical result. The
authors did not discuss these results, but from the respective
observations one could doubt the existence of a yield stress or
expect the non-linear behaviour to be more influential than
the yield stress, since the Bingham and power law model are
both two-parameter models. However, more thorough inves-
tigation would be needed on these hypotheses. In Yang et al.
(2009), the Bingham, Ostwald, Casson and HerscheleBulkley
model were fitted to the data and the latter two were
denominated as the best options according to the R2 and
standard error value. By rewriting the equations, all could be
fitted using a linear regression y ¼ ax þ b, although Her-
scheleBulkley is a three-parameter model. Therefore, s0 was
calculated by extrapolation in the latter, thereby still
providing one extra degree of freedom thus hampering an
Fig. 11 e Illustration of the inappropriateness of only using
R2 as a model selection criterion: both the Bingham and
Ostwald solution provide an equal R2 value, but the data
clearly show a cavity: residuals of the linear model are not
equally distributed (Laera et al., 2007).
entirely fair comparison. In Garakani et al. (2011), seven
models were compared using R2 values as the only selection
criterion, obviously indicating Carreau and Cross models as
the best fitting, followed by HerscheleBulkley and Casson. It
was stated that since the behaviour of a non-Newtonian fluid
includes Newtonian-like behaviour at very low and very high
shear rates, three parameters are necessary to describe the
whole profile accurately. The authors remarked however that,
mostly, the available data were not sufficient for the estima-
tion of all three parameters.
It should be stressed at this point that the use of over-
parameterized models will lead to erroneous conclusions and
decisions, and will eventually lead to a loss of faith in models.
The point is that this is actually caused by the modeller and
not by the model itself. Hence, it is recommended that, before
using a model reported in literature, the procedure used to
derive it should be thoroughly checked. This is especially true
for the data-driven empirical models derived from data that
are heavily dependent on the method in which they were
collected (rheometer device, measurement protocol,.).
Next to a selection of literature models, Yang et al. (2009)
proposed two correlation models including the influence of
MLSS and temperature; one with fitted exponential functions,
one based on the concept of activation energy. Both models
performed equally well based on their R2 value, and it was not
concluded which model to choose. Also, a model accounting for
MLSS and shear rate was provided. This was however built
according to the Power-law equation for pseudoplastic fluids,
while the importance of a yield stress was stressed in the
article. No equation was provided including all three predictors.
Petersen et al. (2008) studied the rheology of a filamentous
fermentation broth, specifically including particle size distribu-
tion, but considered the relationship between rheology on one
side and particle concentration and size on the other as too
complex to be solved in the traditional way. Instead, they
developed a statistical model, based on partial least squares (PLS).
Different combinations of predictors were tried, and it was
concluded that size class data, biomass concentration and the
feed mode as predictors were able to deliver a good model quality.
4.3. Parameter estimation
Parameter estimation aims at finding a “unique” set of model
parameters that is able to describe a set of experimental data.
The word unique is very important in this context. Typically,
optimization algorithms can be used to conduct parameter
estimation. However, a frequent mistake by modellers is that
one just accepts the numbers from the algorithm, without
paying attention to their quality. This is especially problem-
atic when using over parameterized models that suffer from
identifiability problems. In that case, many parameter sets
will result in an equally good fit and, hence, the set is not
unique. Even the most sophisticated optimization algorithm
will not be able to solve this issue as it is inherently related to
the optimization problem. A way to figure out whether this
problem has occurred or not is by checking the parameter
estimation quality (Gujer, 2008). This can be judged from the
confidence intervals of the respective parameter estimates
which can be derived from the parameter estimation var-
ianceecovariance matrix, which can be approximated by the
 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2
inverse of the Fisher Information Matrix (FIM) (Dochain and
Vanrolleghem, 2001). When acceptably small confidence
intervals are found, one can have more trust in the retrieved
estimated parameter set and this will result in a model with
a higher predictive power. Large confidence intervals indicate
identifiability problems. One reason why this is often violated
is that this calculation of confidence intervals is not readily
available in the software used (e.g. Excel solver). It is therefore
good practice to use software that does compute this or
implement a method that allows to estimate confidence
intervals. Although this might seem time consuming, it really
can make a difference between a good and a bad model.
In literature, the proper way of dealing with parameter
estimation is often violated as well. Bailey and Weir (1998) used
the direct parameter estimation method to compare a couple
of rheology models for drilling fluids. This direct parameter
estimation only uses a minimum number of data points
(at discrete shear rates) equal to the number of parameters in
the model, so the result is an equal number of equations, being
a solvable set. The examples given earlier in the model selec-
tion part actually also suffer from this when only using R2. It is
very likely that most of these models will suffer from param-
eter estimates with large confidence intervals.
From the same parameter estimation varianceecovariance
matrix, the correlation between parameters can be derived.
Highly correlated parameters should never be estimated
simultaneously. This is clearly shown in Petersen et al. (2008),
where the HerscheleBulkley model was fitted to the data but
showed highly correlated parameters. Since the flow behaviour
Fig. 12 e Mean and standard deviation of parameter estimates for s0 and k of the HerscheleBulkley model with variable n (a
and b) and fixed n (c and d) (Petersen et al., 2008).
479
index n was found to be insignificantly correlated to any of the
model predictors, its value was fixed, thereby greatly reducing
the uncertainty (i.e. smaller confidence intervals) on the other
two parameters, k and s0 (Fig. 12). However, the problem of
parameter estimation was not solved yet. The estimated values
for k and s0 appeared very sensitive for the chosen (fixed) value
of n. Furthermore, both with fixed or variable n, the (un)
certainty of the model predictions for the apparent viscosity
were not affected by the parameter uncertainty, according to
a bootstrap analysis. This leads to an impasse; one can reduce
uncertainty on parameter estimations by fixing the abundant
parameters, but this will not enhance prediction uncertainty
since the fixed parameter values still exert their influence as
a consequence of the high correlations.
The quality of the parameter estimation depends on the
model structure, the available data sets and their quality. Hence,
the reason for the above-mentioned problem likely resides in
the fact that the collected data do not contain the information
necessary for the current model structures, and that model
structures are not adequate in describing the observations.
4.4. Validation
All of the problems above can be detected when the model is
exposed to a validation data set, and its performance is eval-
uated based on a predefined criterion. This step should not be
forgotten in the modelling process, since it can test the
predictive power of the model and point out flaws for which
one needs to iterate back to previous steps. However, in
480 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2
literature, almost no information can be found on validation
of rheology models. This should be taken into account when
these literature models are used for practical purposes.
5. Conclusion and perspectives
AS rheology is an important property with respect to process
performance and economics. The following key conclusions
and perspectives can be put forward from this review:
 A large number of references about viscosity of AS is avail-
able in literature. Despite this, no consensus has been
reached on how to perform the rheological measurements;
 In the majority of the studied papers, the experimental data
collection was not described in enough detail or not at all;
 In the cases where details of the experiments were given,
a lot of differences can be observed: e.g. choice of rheometer
and exact measurement protocol (shear rate or shear stress
ramp up, interval between ramp, initial flocculation state of
the AS). It is therefore no real surprise that many different
rheological values and models were obtained.
 Repeatability could be improved by better defining how
a rheological measurement of AS should be performed. It
should be noted that this could be different depending on
the purpose of the measurement. Development of such
protocol(s) should be pursued;
 At minimum, authors should mention in detail the rheometer
type and detailed measurement procedure adopted when
publishing rheological measurements of AS;
 Many authors violate different steps of Good Modelling
Practice, especially when performing parameter estimation
and model selection. Model validation is hardly ever per-
formed. This often leads to useless models with low
predictive power, which is dangerous to use. It is therefore
recommended that when using a model from literature one
should always check whether the model was properly
developed;
 Depending on the modelling goal, different model struc-
tures might be required. There is not such a thing as one
optimal model structure for all purposes;
 All current models suffer from parameterization problems,
mainly related to the lack of experimental data and insight
in the property of viscosity of dispersions;
 Knowledge build-up is urgently needed for better process
design and is ideally performed using a model-based
approach in conjunction with dedicated experiments, in
which important factors like particle size distribution are
considered.
references
Aranowski, R., Hupka, J., Jungnickel, C., 2010. Changes in rheological
properties during anaerobic digestion of activated sludge.
Physicochemical Problems of Mineral Processing 44, 13e22.
Bailey, W.J., Weir, I.S., 1998. Investigation of methods for direct
rheological model parameter estimation. Journal of Petroleum
Science and Engineering 21 (1e2), 1e13.
Banu, Ö., 2007. Optimization of a full-scale dewatering operation
based on the rheological characteristics of wastewater sludge.
Water Research 41 (6), 1243e1252.
Barbot, E., Seyssiecq, I., Roche, N., Marrot, B., 2010. Inhibition of
activated sludge respiration by sodium azide addition: effect
on rheology and oxygen transfer. Chemical Engineering
Journal 163 (3), 230e235.
Baudez, J.C., Markis, F., Eshtiaghi, N., Slatter, P., 2011. The
rheological behaviour of anaerobic digested sludge. Water
Research 45 (17), 5675e5680.
Belia, E., Amerlinck, Y., Benedetti, L., Johnson, B., Sin, G.,
Vanrolleghem, P.A., Gernaey, K.V., Gillot, S., Neumann, M.B.,
Rieger, L., Shaw, A., Villez, K., 2009. Wastewater treatment
modelling: dealing with uncertainties. Water Science and
Technology 60 (8), 1929e1941.
Benedetti, L., Batstone, D.J., De Baets, B., Nopens, I.,
Vanrolleghem, P.A., 2012. Uncertainty analysis of WWTP
control strategies made feasible. Water Quality Research
Journal of Canada 47 (1), 14e29.
Biggs, C., Lant, P., Hounslow, M., 2003. Modelling the effect of
shear history on activated sludge flocculation. Water Science
and Technology 47 (11), 251e257.
Bird, R.B., Stewart, W.E., Lightfoot, E.N., 2001. Transport
Phenomena. Wiley.
Brannock, M., Leslie, G., Wang, Y., Buetehorn, S., 2010a.
Optimising mixing and nutrient removal in membrane
bioreactors: CFD modelling and experimental validation.
Desalination 250 (2), 815e818.
Brannock, M., Wang, Y., Leslie, G., 2010b. Mixing characterisation
of full-scale membrane bioreactors: CFD modelling with
experimental validation. Water Research 44 (10), 3181e3191.
Bürger, R., Diehl, S., Nopens, I., 2011. A consistent modelling
methodology for secondary settling tanks in wastewater
treatment. Water Research 45 (6), 2247e2260.
Campbell, H.W., Crescuolo, P.J., 1982. Use of rheology for sludge
characterization. Water Science and Technology 14 (6),
475e489.
Chen, B., Lee, S., Lee, D.J., 2005. Rheological characteristics of the
cationic polyelectrolyte flocculated wastewater sludge. Water
Research 39 (18), 4429e4435.
Christopher, F.F., 2002. The rheological and physico-chemical
characteristics of sewage sludges. Enzyme and Microbial
Technology 30 (3), 340e345.
Cierkens, K., Plano, S., Benedetti, L., Weijers, S., de Jonge, J.,
Nopens, I., 2012. Impact of influent data frequency and
model structure on the quality of WWTP model calibration
and uncertainty. Water Science and Technology 65 (2),
233e242.
Cornel, P., Wagner, M., Krause, S., 2003. Investigation of oxygen
transfer rates in full scale membrane bioreactors. Water
Science and Technology 47 (11), 313e319.
De Clercq, B., 2003. Computational Fluid Dynamics of Settling
Tanks: Development of Experiments and Rheological, Settling
and Scraper Submodels. Ghent University.
Defrance, L., Jaffrin, M.Y., Gupta, B., Paullier, P., Geaugey, V., 2000.
Contribution of various constituents of activated sludge to
membrane bioreactor fouling. Bioresource Technology 73 (2),
105e112.
Dick, R.I., Ewing, B.B., 1967. The rheology of activated sludge.
Journal (Water Pollution Control Federation) 39 (4), 543e560.
Dochain, D., Vanrolleghem, P., 2001. Dynamical Modelling and
Estimation in Wastewater Treatment Processes. IWA
Publishing, London.
Dochain, D., Vanrolleghem, P.A., Van Daele, M., 1995. Structural
identifiability of biokinetic models of activated sludge
respiration. Water Research 29, 2571e2578.
Eshtiaghi, N., Markis, F., Slatter, P., 2012. The laminar/turbulent
transition in a sludge pipeline. Water Science and Technology:
 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2
A Journal of the International Association on Water Pollution
Research 65 (4), 697e702.
Fabiyi, M.E., Novak, R., 2008. Evaluation of the factors that impact
successful membrane biological reactor operations at high
solids concentration. Proceedings of the Water Environment
Federation 2008 (1), 503e512.
Garakani, A.H.K., Mostoufi, N., Sadeghi, F., Hosseinzadeh, M.,
Fatourechi, H., Sarrafzadeh, M.H., Mehrnia, M.R., 2011.
Comparison between different models for rheological
characterization of activated sludge. Iranian Journal of
Environmental Health Science & Engineering 8 (3), 255e264.
Germain, E., Nelles, F., Drews, A., Pearce, P., Kraume, M., Reid, E.,
Judd, S.J., Stephenson, T., 2007. Biomass effects on oxygen
transfer in membrane bioreactors. Water Research 41 (5),
1038e1044.
Gil, J.A., Krzeminski, P., van Lier, J.B., van der Graaf, J.H.J.M.,
Wijffels, T., Prats, D., 2011. Analysis of the filterability in
industrial MBRs. Influence of activated sludge parameters and
constituents on filterability. Journal of Membrane Science
385e386 (0), 96e109.
Gujer, W., 2008. Systems Analysis for Water Technology. Springer
Verlag, Berlin.
Guibaud, G., Dollet, P., Tixier, N., Dagot, C., Baudu, M., 2004.
Characterisation of the evolution of activated sludges using
rheological measurements. Process Biochemistry 39 (11),
1803e1810.
Guibaud, G., Tixier, N., Baudu, M., 2005. Hysteresis area, a rheological
parameter used as a tool to assess the ability of filamentous
sludges to settle. Process Biochemistry 40 (8), 2671e2676.
Hammadi, L., Ponton, A., Belhadri, M., 2011. Rheological study
and valorization of waste sludge from wastewater treatment
plants in the dredging operation of hydraulic dams. Energy
Procedia 6 (0), 302e309.
Hasar, H., Kinaci, C., Ünlü, A., Tog rul, H., Ipek, U., 2004.
Rheological properties of activated sludge in a sMBR.
Biochemical Engineering Journal 20 (1), 1e6.
Jin, B., Yin, P., Lant, P., 2006. Hydrodynamics and mass transfer
coefficient in three-phase air-lift reactors containing activated
sludge. Chemical Engineering and Processing: Process
Intensification 45 (7), 608e617.
Jones, W.M., Rees, I.J., 1982. The stringiness of dilute polymer-
solutions. Journal of Non Newtonian Fluid Mechanics 11 (3e4),
257e268.
Kornboonraksa, T., Lee, S.H., 2009. Factors affecting the
performance of membrane bioreactor for piggery wastewater
treatment. Bioresource Technology 100 (12), 2926e2932.
Laera, G., Giordano, C., Pollice, A., Saturno, D., Mininni, G., 2007.
Membrane bioreactor sludge rheology at different solid
retention times. Water Research 41 (18), 4197e4203.
Li, J., Yang, F., Li, Y., Wong, F., Chua, H.C., 2008. Impact of
biological constituents and properties of activated sludge on
membrane fouling in a novel submerged membrane
bioreactor. Desalination 225 (1e3), 356e365.
Linstrom, P.J, Mallard, W., 2009. NIST Chemistry WebBook, NIST
Standard Reference Database Number 69.
Marinetti, M., Dentel, S.K., Malpei, F., Bonomo, L., 2010.
Assessment of rheological methods for a correlation to sludge
filterability. Water Research 44 (18), 5398e5406.
Meng, F., Shi, B., Yang, F., Zhang, H., 2007. New insights into
membrane fouling in submerged membrane bioreactor based
on rheology and hydrodynamics concepts. Journal of
Membrane Science 302 (1e2), 87e94.
Metzner, A.B., Otto, R.E., 1957. Agitation of non-Newtonian fluids.
AIChE Journal 3 (1), 3e10.
Mikkelsen, L.M., 2001. The shear sensitivity of activated sludge:
relations to filterability, rheology and surface chemistry.
Colloids and Surfaces A Physicochemical and Engineering
Aspects 182 (1e3), 1e14.
481
Moeller, G., Torres, L.G., 1997. Rheological characterization of
primary and secondary sludges treated by both aerobic and
anaerobic digestion. Bioresource Technology 61 (3),
207e211.
Moreau, A.A., Ratkovich, N., Nopens, I., van de Graff, J., 2009. The
(in)significance of apparent viscosity in full-scale municipal
membrane bioreactors. Journal of Membrane Science 340
(1e2), 249e256.
Mori, M., Seyssiecq, I., Roche, N., 2006. Rheological measurements
of sewage sludge for various solids concentrations and
geometry. Process Biochemistry 41 (7), 1656e1662.
Nopens, I., Biggs, C.A., De Clercq, B., Govoreanu, R.,
Wilen, B.M., Lant, P., Vanrolleghem, P.A., 2002. Modelling
the activated sludge flocculation process combining laser
light diffraction particle sizing and population balance
modelling (PBM). Water Science and Technology 45 (6),
41e49.
Nopens, I., Sin, G., Jiang, T., d’Antonio, L., Stama, S., Zhao, J.,
Vanrolleghem, P.A., 2007. Model-based optimisation of the
biological performance of a sidestream MBR. Water Science
and Technology 56 (6), 135e143.
Oliver, D., Bragg, R., 1974. The triple jet: a new method for
measurement of extensional viscosity. Rheologica Acta 13 (4),
830e835.
Paulo Santos, M., 1997. The influence of the anaerobic digestion
process on the sewage sludges rheological behaviour. Water
Science and Technology 36 (11), 61e67.
Petersen, N., Stocks, S., Gernaey, K.V., 2008. Multivariate models
for prediction of rheological characteristics of filamentous
fermentation broth from the size distribution. Biotechnology
and Bioengineering 100 (1), 61e71.
Pevere, A., Guibaud, G., van Hullebusch, E., Lens, P., 2007.
Identification of rheological parameters describing the
physico-chemical properties of anaerobic sulphidogenic
sludge suspensions. Enzyme and Microbial Technology 40 (4),
547e554.
Pevere, A., Guibaud, G., van Hullebusch, E., Lens, P., Baudu, M.,
2006. Viscosity evolution of anaerobic granular sludge.
Biochemical Engineering Journal 27 (3), 315e322.
Pollice, A., Giordano, C., Laera, G., Saturno, D., Mininni, G., 2007.
Physical characteristics of the sludge in a complete retention
membrane bioreactor. Water Research 41 (8), 1832e1840.
Pollice, A., Giordano, C., Laera, G., Saturno, D., Mininni, G., 2006.
Rheology of sludge in a complete retention membrane
bioreactor. Environmental Technology 27 (7), 723e732.
Pollice, A., Laera, G., Saturno, D., Giordano, C., 2008. Effects of
sludge retention time on the performance of a membrane
bioreactor treating municipal sewage. Journal of Membrane
Science 317 (1e2), 65e70.
Popovic, M., Robinson, C. W., 1984. Estimation of some important
design parameters for non-Newtonian liquids in
pneumatically-agitated fermenters. In: 34th Canadian
Chemical Engineering Congress., pp. 258e263.
Proff, E.A., Lohmann, J.H., 1997. Calculation of pressure drop in
the tube flow of sewage sludges with the aid of flow curves.
Water Science and Technology 36 (11), 27e32.
Rios, N., Nopens, I., Jiang, T., De Shepper, V., Jiang, T., Verstraete
W., Vanrolleghem P., 2007. A rheological model for activated
sludge in a Side-Stream MBR. In: Proceedings of the IWA
membrane conference on May 15e17 2007 in Harrogate, UK.
Rosenberger, R., Kubin, K., Kraume, M., 2002. Rheology of
activated sludge in membrane bioreactors. Engineering in Life
Sciences 2 (9), 269e275.
Saltelli, A., Chan, K., Scott, E.M., 2000. Sensitivity Analysis. In:
Wileys Series in Probability and Statistics. Wiley, New York.
Sanin, F.D., 2002. Effect of solution physical chemistry on the
rheological properties of activated sludge. Water SA 28 (2),
207e211.
482 w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 4 6 3 e4 8 2
Schumacher, S., 2006. Leistungsbestimmende Prozesse in
Nachklärbecken. Einflussgrößen, Modellbildung und
Optimierung. PhD thesis., Hannover University.
Seyssiecq, I., Marrot, B., Djerroud, D., Roche, N., 2008. In situ
triphasic rheological characterisation of activated sludge, in an
aerated bioreactor. Chemical Engineering Journal 142 (1), 40e47.
Seyssiecq, I., Ferrasse, J.H., Roche, N., 2003. State-of-the-art:
rheological characterisation of wastewater treatment sludge.
Biochemical Engineering Journal 16 (1), 41e56.
Slatter, P., 2001. Sludge pipeline design. Water Science and
Technology 44 (10), 367. VII.
Slatter, P., 1997. Rheological characterisation of sludges. Water
Science and Technology 36 (11), 9e18.
Slatter, P., 2004. The hydraulic transportation of thickened
sludges. Water SA 30 (5), 614e616.
Spinosa, L., Lotito, V., 2003. A simple method for evaluating sludge
yield stress. Advances in Environmental Research 7 (3), 655e659.
Tchobanoglous, G., Burton, F.L., Stensel, H.D., 2003. Wastewater
Engineering: Treatment and Reuse. McGraw-Hill, Boston.
Tixier, N., Guibaud, G., Baudu, M., 2003a. Determination of some
rheological parameters for the characterization of activated
sludge. Bioresource Technology 90 (2), 215e220.
Tixier, N., Guibaud, G., Baudu, M., 2003b. Towards a rheological
parameter for activated sludge bulking characterisation.
Enzyme and Microbial Technology 33 (2e3), 292e298.
Usui, H., Sano, Y., 1981. Elongational flow of dilute drag reducing
fluids in a falling jet. Physics of Fluids 24 (2), 214e219.
Van Kaam, R., Anne-Archard, D., Gaubert, M.A., Albasi, C., 2008.
Rheological characterization of mixed liquor in a submerged
membrane bioreactor: interest for process management.
Journal of Membrane Science 317 (1e2), 26e33.
Wang, Y.L., Dentel, S.K., 2010. The effect of high speed mixing and
polymer dosing rates on the geometric and rheological
characteristics of conditioned anaerobic digested sludge
(ADS). Water Research 44 (20), 6041e6052.
Wu, Z., Wang, Z., Zhou, Z., Yu, G., Gu, G., 2007. Sludge rheological
and physiological characteristics in a pilot-scale submerged
membrane bioreactor. Desalination 212 (1e3), 152e164.
Xia, M., Wang, Z., Wu, Z., Wang, X., Zhou, Z., Lu, J., 2009.
Simulation and assessment of sludge concentration and
rheology in the process of waste activated sludge treatment.
Journal of Environmental Sciences 21 (12), 1639e1645.
Yang, F., Bick, A., Shandalov, S., Brenner, A., Oron, G., 2009. Yield
stress and rheological characteristics of activated sludge in an
airlift membrane bioreactor. Journal of Membrane Science 334
(1e2), 83e90.
Yen, P., Chen, L.C., Chien, C.Y., Wu, R., Lee, D.J., 2002. Network
strength and dewaterability of flocculated activated sludge.
Water Research 36 (3), 539e550.
Zhao, W., Huang, X., Lee, D., Wang, X., Shen, Y., 2009. Use of
submerged anaerobiceanoxiceoxic membrane bioreactor
to treat highly toxic coke wastewater with complete
sludge retention. Journal of Membrane Science 330 (1e2),
57e64.